Optimization of neural network architecture using genetic algorithms: a case study in diabetes prediction

The article presents the research project GANet, which focuses on the optimization of neural network architectures for diabetes prediction using genetic algorithms. The main goal of the project is to develop and analyze the effectiveness of a genetic algorithm that would facilitate the automation and optimization of the design of neural network architectures for binary classification tasks. A distinctive feature of GANet is an innovative mutation strategy that allows neurons to mutate within sub-networks with varying numbers of layers and neurons, thereby contributing to the evolution of more complex and efficient structures. Experiments were conducted on a publicly available diabetes dataset, and the results demonstrate a significant improvement in the accuracy of diabetes prediction. The study confirms the effectiveness of using genetic algorithms for optimizing neural network architectures and opens new prospects for future research in this field.